Local Histogram Equalization in PI.

[ollypenrice]

Encouraged by Harry Page I've been trying this routine. I like it as a local contrast enhancer, though I saved it as an image to use in a layer in Photoshop (sorry Harry!!) so I could control where I wanted it. I'm wary of heading into the hard look that PI encourages so I don't know what you think of the before and after.

Before;  http://ollypenrice.smugmug.com/Other/Best-of-Les-Granges/i-cfM37HH/0/X3/CONE%2085%20106%20HaRGB-X3.jpg

After;    http://ollypenrice.smugmug.com/Other/Best-of-Les-Granges/i-tKmhtWg/0/O/CONE%20HaLRGB.%20Version%202..jpg

Olly

[johnrt]

Olly, I do 95% of my processing in Pixinsight and consciously work to avoid the "Processed in PI" look of overbearing wavelets and contrast enhancements. I strive to keep a softer look but with good contrast in the parts where it's needed. To do this I use the local histo equalisation tool. It's powerful indeed but a key part of my processing and excellent when used in moderation.

When I use it I build a mask using a tutorial video I found from a PI Jedi to mask off everything I don't want enhanced, the photoshop layer method will achieve the same result, and dial back the slider to 50% or lower of the enhanced version to be blended.

In conclusion it gets a thumbs up from me.

[JohnC64]

I opened both up in separate tabs in a browser and then toggled between the two.  The second one definitely has more contrast, but from a purely  personal aesthetic point of view, i prefer the first one as it has a more natural look to it.

johnrt - do you have a link to the tutorial video.   Always interested in watching a PI Jedi at work. :-)

John

[ollypenrice]

Olly, I do 95% of my processing in Pixinsight and consciously work to avoid the "Processed in PI" look of overbearing wavelets and contrast enhancements. I strive to keep a softer look but with good contrast in the parts where it's needed. To do this I use the local histo equalisation tool. It's powerful indeed but a key part of my processing and excellent when used in moderation.

When I use it I build a mask using a tutorial video I found from a PI Jedi to mask off everything I don't want enhanced, the photoshop layer method will achieve the same result, and dial back the slider to 50% or lower of the enhanced version to be blended.

In conclusion it gets a thumbs up from me.

Your processing is absolutely the opposite of what I don't like in hardcore PI so it can clearly be done. It's some of the most gentle and invisible on the net in my view. Interesting.

Olly

[harry page]

Hi Olly

I use a mask to protect the areas I don't want LHE but what ever floats your boat with PS

But I am glad you have found the tool as I believe it is very powerful and controllable

we could have a discussion for months on end about whats right is astrophotography , but again get your boat out

Regards

harry

[johnrt]

johnrt - do you have a link to the tutorial video.   Always interested in watching a PI Jedi at work. :-)

John

The Jedi in question is Gerald Wechselberger, his tutorials page is at http://www.werbeagentur.org/oldwexi/PixInsight/PixInsight.html be prepared for some mind blowing pixelmath equations!!

[Physicist13]

PI: Image Processing

PS: Artistry

[IanL]

Second version for me, the true 3D structure is far more apparent whereas the first version looks a bit flat.  I agree that some PI processing takes things way too far and you can end up with some very unnatural looking images, but this certainly doesn't have that feel.

[ollypenrice]

PI: Image Processing

PS: Artistry

Sorry but I regard this as totally untrue. There is nothing remotely scientific about PI. It is a digital graphics programme like Photoshop. It works on modified histograms rather than linear ones and therefore none of the brightnesses in a final image are those of the captured data. It has many processing tools which read the measured values of the original and modify them in order to emphasize differences latent in the data. The manner in which it does so is controlled by the user who sets limits on the spacial and numerical range of those alterations based on personal taste. This describes noise reduction, sharpening, DBE and ABE, Wavelets, Local Histogram Equalization, Enhance Dark Structures, etc etc etc. It might be easier in Photoshop to invent things. I can write 'Ceci n'est pas une pipe' on my image more easily in Photoshop than in PI, perhaps, but I don't do that. In processing an image in Ps I work, as does PI, by emphasizing what is already latent in the data.

On what basis do you consider PI to be different in kind from Photoshop? In my view they are remarkably similar, the main difference that PI's selection tools are imposed on the image by a mathematical UI while Photoshop's can be done this way or they can be done manually. If they are done manually then they can be done without regard to the captured data but does anybody do that?  There is a famous image, PI propcessed, of NGC1333 in which some blue reflection nebulosity has a hard boundary. I have the same object captured  in the same telescope and I can assure you that this boundary simply does not exist. You can invent and distort in either PI or Ps. Good imagers try not to do so.

Olly

[harry page]

Hi Olly

Well I am a bit more In the middle than you

I do believe that philosophy in Pi is more Data related and I don't personally like the Painting that goes on in Photoshop even if that is brushing in things through masks

and just selecting things because you think you should

As I always say this is my opinion and it takes all sorts even photoshop users

Keep well

Harry the Hard Man processor

[ollypenrice]

Essentailly I think we agree, Harry. The processing should be data related and not painted. I also think you can paint in Photoshop, as I've said above. But that doesn't mean you have to or that many people do.

Let's compare ways of selecting what to noise reduce. It will be areas of faint signal, I'm sure we'll both agree. In PI you will mask the brighter areas aand apply a variant of a blur to what is not masked. In Ps I will select by colour range the faint signal and apply the NR blur to that. I can see no philosophical difference whatever between these selection methods. In both cases it is the imager who makes the selection, aided by the mathematicians who wrote the masking/selecting tools.

Or sharpening. Nobody wants to sharpen everything, they want to select the areas for which sharpening is appropriate. In PI you can mask off the stars and the faint signal and sharpen. Likewise in Ps, or you can use layers and erase, with a brush, the unsharpened regions to taste. But the brush isn't a paintbrush, it is a selection tool.

HDR wavelets or Layer Masking? Here I'm certain that layer masking in Ps is more accurate than PI wavelets. When I take a short set of subs for the Trapezium and a long set, and apply a log stretch to both, I can see what both look like and I blend them, keeping the original look. Wavelets, though, create a new look not seen in the orginal data, short or long.

I like both programmes, let me stress. I just don't believe there is any significant philosophical difference between them.

Olly

[wxsatuser]

Second version for me, the true 3D structure is far more apparent whereas the first version looks a bit flat.  I agree that some PI processing takes things way too far and you can end up with some very unnatural looking images, but this certainly doesn't have that feel.

I would go along with Ian, although a newbie and have'nt a clue, yet!

Mulled over both for quite a while and would find both very pleasing if shown seperately.

I think I just prefer the 2nd as Ian above, it does'nt look over done and the highlights give it the edge.

If I could produce either of these images I would be one happy bunny.

[RikM]

The new one certainly has more pazazz and I like it a lot but could you not achieve the same result with a masked curves stretch and LAB colour or Match colour boost 

[ollypenrice]

The new one certainly has more pazazz and I like it a lot but could you not achieve the same result with a masked curves stretch and LAB colour or Match colour boost 

Maybe so, yes. I found LHE brought out more colour contrasts than the Lab method, which had already been applied to the first one.

Olly

[RikM]

LHE has definitely done the job. My only negative would be the bright little red knot upper-left. On my monitor it looks a wee bit too much and sticks out rather than stands out if you know what I mean. But that could equally well be my monitor.

[ollypenrice]

I know what you mean. It looks OK on mine but, equally, that could be mine! I liked the way LHE distiguished it from the magentas, though. Overdone, maybe, but OK in principle, I reckon.

Olly

[Jannis]

I love the "after" version much better. Much stronger and brighter colors combined with better contrast.

While both are good, teh later one look much more "alive!"

[harry page]

Essentailly I think we agree, Harry. The processing should be data related and not painted. I also think you can paint in Photoshop, as I've said above. But that doesn't mean you have to or that many people do.

Let's compare ways of selecting what to noise reduce. It will be areas of faint signal, I'm sure we'll both agree. In PI you will mask the brighter areas aand apply a variant of a blur to what is not masked. In Ps I will select by colour range the faint signal and apply the NR blur to that. I can see no philosophical difference whatever between these selection methods. In both cases it is the imager who makes the selection, aided by the mathematicians who wrote the masking/selecting tools.

Or sharpening. Nobody wants to sharpen everything, they want to select the areas for which sharpening is appropriate. In PI you can mask off the stars and the faint signal and sharpen. Likewise in Ps, or you can use layers and erase, with a brush, the unsharpened regions to taste. But the brush isn't a paintbrush, it is a selection tool.

HDR wavelets or Layer Masking? Here I'm certain that layer masking in Ps is more accurate than PI wavelets. When I take a short set of subs for the Trapezium and a long set, and apply a log stretch to both, I can see what both look like and I blend them, keeping the original look. Wavelets, though, create a new look not seen in the orginal data, short or long.

I like both programmes, let me stress. I just don't believe there is any significant philosophical difference between them.

Olly

Hi Olly

You can do HDR combination in pixinsight with different subs very easy in deed give it a go some time   http://harrysastroshed.com/HDR%20comp.html

Harry

[jnc71106]

Hi Olly

You can do HDR combination in pixinsight with different subs very easy in deed give it a go some time   http://harrysastroshed.com/HDR%20comp.html

Harry

"who said pixinsight was difficult" that did make me chuckle